Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]


Groups > comp.soft-sys.math.mathematica > #6522 > unrolled thread

Timing function

Started byMark <big.dog1@att.net>
First post2011-11-19 11:46 +0000
Last post2011-11-21 09:38 +0000
Articles 3 — 3 participants

Back to article view | Back to comp.soft-sys.math.mathematica


Contents

  Timing function Mark <big.dog1@att.net> - 2011-11-19 11:46 +0000
    Re: Timing function "Oleksandr Rasputinov" <oleksandr_rasputinov@hmamail.com> - 2011-11-20 10:39 +0000
    Re: Timing function David Bailey <dave@removedbailey.co.uk> - 2011-11-21 09:38 +0000

#6522 — Timing function

FromMark <big.dog1@att.net>
Date2011-11-19 11:46 +0000
SubjectTiming function
Message-ID<ja84v7$ifg$1@smc.vnet.net>
Hello community:

When I run the following expression below, It returns the value of zero for the timing. If I press "shift enter" again, it then correctly returns the amount of time for the expression to run / plot.

Also, I noticed that if I change the variable from x^3 to x^2 and press "shift enter" at that point, zero again is returned for the timing result. I am using Mathematica 8.01

Timing[Plot[x^3, {x, -5, 5}]]

Does anyone know the reason for this?

[toc] | [next] | [standalone]


#6553

From"Oleksandr Rasputinov" <oleksandr_rasputinov@hmamail.com>
Date2011-11-20 10:39 +0000
Message-ID<jaald1$17h$1@smc.vnet.net>
In reply to#6522
On Sat, 19 Nov 2011 11:46:47 -0000, Mark <big.dog1@att.net> wrote:

> Hello community:
>
> When I run the following expression below, It returns the value of zero  
> for the timing. If I press "shift enter" again, it then correctly  
> returns the amount of time for the expression to run / plot.
>
> Also, I noticed that if I change the variable from x^3 to x^2 and press  
> "shift enter" at that point, zero again is returned for the timing  
> result. I am using Mathematica 8.01
>
> Timing[Plot[x^3, {x, -5, 5}]]
>
> Does anyone know the reason for this?
>

This observation is probably due to clock granularity. On Windows, for  
example, the system clock has a (default) resolution of 1/64s; each clock  
tick triggers an interrupt and consumes some CPU time so it is not helpful  
to have an extremely high resolution system clock. As a result (again, on  
Windows) the maximum supported resolution is 0.5ms. Since plotting a  
simple function takes only a short time, it is a matter of chance as to  
whether the timing is reported as 0 or 1 clock ticks: this will be  
reported as either 0. or 0.015625 seconds in the output from Timing. Other  
systems use different clock granularities, but in general one should not  
trust timings that are at the limit of the system's ability to measure.  
Instead, either repeat the operation sufficiently many times (e.g. using  
Do) that the total timing is much longer than a single clock tick, or use  
a timing mechanism with higher resolution than the system clock.

[toc] | [prev] | [next] | [standalone]


#6597

FromDavid Bailey <dave@removedbailey.co.uk>
Date2011-11-21 09:38 +0000
Message-ID<jad65o$ek8$1@smc.vnet.net>
In reply to#6522
On 19/11/2011 11:46, Mark wrote:
> Hello community:
>
> When I run the following expression below, It returns the value of zero for the timing. If I press "shift enter" again, it then correctly returns the amount of time for the expression to run / plot.
>
> Also, I noticed that if I change the variable from x^3 to x^2 and press "shift enter" at that point, zero again is returned for the timing result. I am using Mathematica 8.01
>
> Timing[Plot[x^3, {x, -5, 5}]]
>
> Does anyone know the reason for this?
>

The clock granularity has already been mentioned, but I suspect perhaps 
that if you are timing Plot commands, you are struggling with more 
complex plots that are irritatingly slow! A complication is that Timing 
only measures the time used by the kernel, and for some plots, most of 
the time is consumed in the frontend. Even AbsoluteTime does not solve 
this problem easily, because it returns its result before the FE starts 
its work.

I gave a talk at the 2010 autumn conference in Champaign discussing a 
variety of subtle timing issues including this problem. You will find a 
link to this presentation here:

http://www.wolfram.com/events/techconf2010/speakers.html

David Bailey
http://www.dbaileyconsultancy.co.uk

[toc] | [prev] | [standalone]


Back to top | Article view | comp.soft-sys.math.mathematica


csiph-web